VEHICLE CLASSIFIER FOR BRIDGE RATING MEASUREMENT SYSTEM USING WIRELESS SENSOR NETWORK

• Main Author: ANANDADIGA (NIM: 13214086), ABIKARAMI
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/24922
• Institution: Institut Teknologi Bandung

<p align="justify"> <br /> <br /> <br /> <br /> Bridge is one of the most important infrastructure to connect two places. Currently, in Indonesia, to monitor bridge’s structural health generally it still uses manual method on yearly time basis, so latest information of the structural health could not be known when it is not the time for inspection. Therefore, Bridge Rating Monitoring System using Wireless Sensor Network is proposed to solve monitoring problem. This system is designed to measure the rating and mode shape of the bridge due to vibration generated by particular vehicle crossing the bridge. This rating and mode shape is expected to be able to be used as early diagnosis of the bridge health. The whole system consists of Activator sub-systems as sensor activation controllers, Sensory Data Processing for data acquisition and processing, and Server & GUI to store data and to display information to users. In this book, discussed subject is only Vehicle Classifier which is part of the Activator. Classifier is used to classify the class of the passing trucks, as one of the parameter data for the Activator in order to determine whether the sensor node is activated. Classifier uses Sun Microsystems’ SunSPOT as hardware. Classifier is designed to group the trucks passing into three classes by their axle – 2 axles, 3 axles, and 4 or more axles –, and using Artificial Neural Network as Neural Network (NN) prediction model architecture. Input for the NN are the results of FFT from bridge vibration signal, that is amplitude displacement peak and its corresponding frequency. NN training can be done offline, so that for the prediction model in SunSPOT we only need to implement the value of weight and bias from training. Based on the test results, the system has pretty well classified the truck group. For future work, datasets on truck weights can be tried as inputs in order to create a better classifier. <p align="justify">